October 18, 2013


Ocean Color POLDER 2 advanced algorithms

Calibration adjustment

We rely on a vicarious calibration that compared POLDER measured radiances data with computed radiances at the top of the atmosphere (TOA) over sites where in-situ measurements of water leaving radiance/reflectance and aerosol optical thickness have been made. Level-1 radiances are adjusted by a few percent at 443 and 490 nm. The accuracy of this vicarious calibration method is assessed to be about 1 % (Fougnie et al., 1998).

More details about POLDER calibration process can be found at:


Atmospheric correction

Gaseous absorption and stratospheric aerosols corrections are made just after clouds masking and before any atmospheric corrections (see Flowchart). The corrected signal can then be decomposed as:

accounts for multiple scattering by molecules, accounts for coupling terms between molecules and photons reflected on a rough sea surface, accounts for multiple scatterings by aerosols, accounts for direct sun-glint and the last term accounts for diffuse contribution from sea-surface, where and are the total downward and upward transmission and the spherical albedo of the atmosphere respectively, and and are the water reflectance and the foam reflectance respectively.

The n directions of the multi-angular view of a same pixel are filtered to reject sun glint contamination.

Atmospheric contributions are computed and archived in Look-Up Tables (LUT). Linear interpolation into these tables are carried on for each pixel along with an adjustment to account for the actual sea level atmospheric pressure.

Combining all directions, the spectral dependence of the aerosol scattering is obtained from the POLDER measurements at 865, 765 and 670 nm and is compared with the spectral dependence of aerosols models obtained from pre-computed tables. The POLDER aerosol models have a definition similar to the ones used for SeaWiFS. Namely the 12 models are: M98, M95, M90, M80, C90, C80, C70, T99, T98, T90, T80, T70, where M, C and T design respectively Maritime, Coastal and Tropospheric models and the 2-digits number is the relative humidity, according to the Shettle & Fenn classification. In order to account for marine reflectance at 670 nm, a straightforward correction has been implemented based on empirical relationship between marine reflectance at 565 and 670 nm.
The aerosol optical thickness is derived from the measurements at 865 nm, knowing the aerosol model. Phase functions and spectral dependencies of those 12 models are plotted figure 1.

Figure 1. Phase function and spectral dependence (using single scattering approximation) of the 12 models used in the algorithm.

Once the aerosol model and optical thickness are known, the atmospheric scattering is computed at 443, 490 and 565 nm, using pre-computed scattering tables, and subtracted from the POLDER measurements at these wavelengths for all directions. The residual is divided by the computed total transmission of the atmosphere and is called directional marine reflectance, given in the level 2 directional product (OC2A).

More details about POLDER atmospheric correction process can be found here in pdf format

Bio-optical algorithm

Five bio-optical parameters are retrieved from the spectral marine reflectances:

    The chlorophyll concentration as estimated from SeaWiFS-OC4 algorithm: Chl1
    The chlorophyll concentration estimated from a new empirical algorithm: Chl2
    The absorption coefficients at 443 and 490 nm: a(443) and a(490)
    The backscattering coefficient at 565 nm: bb(565)

    Chl1, the chlorophyll concentration estimated by a SeaWiFS-like bio-optical algorithm, OC2v4 (O'Reilly et al., 2000), using the ratio of the POLDER derived marine reflectances at 490 and 565 nm. Chl1 is the same product than the previously delivered by POLDER 1.

    Chl2, the chlorophyll concentration estimated with a 3 wavelengths biooptical algorithm, at 443, 490 and 565 nm, customized for POLDER data (Deschamps, 2003). The use of the marine reflectance at 443 nm allows to increase the sensitivity of the biooptical algorithm at low chlorophyll concentration, together with the quadratic combination of channels. Chl2 should be more accurate and should be used preferably to Chl1 after its validation. The two algorithms tend to the same estimate for high chlorophyll concentration.

    Compared to Chl2, a and bb provide new additional biogeochemical information such as particulate load (through bb), and the opportunity to distinguished different families of suspended marine particles (through bb(565)/a(443)). The bio-optical algorithm combines the reflectance at different wavelengths to estimate the chlorophyll-a concentration in the surface layer of the ocean. The retrieval of the absorption and the backscattering coefficients from the POLDER marine spectral reflectance is based on the inverse method developed by Loisel and Stramski [2000]. Note, that prior to retrieve these parameters, the directional marine reflectance w has to be corrected from the directional effects related to the geometry of observation.

    A threshold of 0.5 on the value of the aerosol optical thickness at 865 nm is used to mask highly contaminated observations. A threshold on the marine reflectance at 565 nm is used to classify the water type, Case-1 or Case-2 waters.

    More details about bio-optical algorithms and their results can be found here in pdf format

    Level-3 composite

    The level-3 composite is the mean of parameters obtained from POLDER for a given pixel over a ten-day period (1st-10th, 11th-20th, or 21st-last of month) and during a month. A threshold of 0.3 on the aerosol optical thickness is used to select the best data. A weight which is function of the estimated quality index of the processing also introduced on the daily observation. Arithmetic mean is used for marine reflectance and aerosol parameters and geometric mean is used for chlorophyll and bio-optical parameters. Case-2 pixels are discarded for chlorophyll synthesis.

    © CNES Copyrights 2005